MarketintelligenceplatformforBritishAmericanTobacco
Human analysts physically cover only a small fraction of the market and refresh data once every few weeks — a competitor ships a new product, and the company learns about it late, losing shelf space and share. Scraping covers 15,000+ websites with ~95% attribute recognition and <24 h refresh, saving the payroll of ~12 analysts (≈$180–250k/yr) and removing the market’s blind spots.
A product market intelligence platform — automated collection and analysis of data on devices, flavors, news and social media. After launch the team runs with 2 people instead of 14, with 15,000+ websites and 5,500 Instagram accounts under daily monitoring.
The cost of manual work
Human analysts physically cover only a small fraction of the market and refresh data once every few weeks — a competitor ships a new product, and the company learns about it late, losing shelf space and share.
What AI changes
Scraping covers 15,000+ websites with ~95% attribute recognition and <24 h refresh, saving the payroll of ~12 analysts (≈$180–250k/yr) and removing the market’s blind spots.
Operating model
- 14 analysts monitor the market by hand
- Coverage limited to a small share of sources
- Reports are slow, data goes stale
- 14 → 2 people on the team
- 15,000+ websites under daily monitoring
- 5,500 Instagram accounts under daily monitoring
How we built it
Diagnostic
We mapped how the team monitored the market manually and which sources fell outside coverage. We locked the outcome metrics: source coverage and how fast a new product lands in the base.
Prototype on real data
We built the first working loop of scraping and attribute recognition on real sources — within 1–6 weeks from kickoff to usable data.
Production & support
We scaled coverage to 15,000+ websites and 5,500 Instagram accounts, holding ~95% attribute recognition and <24 h refresh as live quality metrics.
What came out
Igor Golikov
Vitaly Kust
Wingle data platform for the wellness market
28 analysts manually consolidate hundreds of sources — some launches and trends slip through, reports go stale before publication, and manual entry errors distort decisions. The platform keeps 9,000+ websites and 7,000+ accounts in real time and produces a report in minutes instead of weeks, freeing up ~24 positions (≈$150k+/yr in savings).
